MySQL Server 5.6 defaults changes

We're improving the MySQL Server defaults, as announced by Tomas Ulin at MySQL Connect. Here's what we're changing:

 Old  New  Notes
back_log  50  50 + ( max_connections / 5 ) capped at 900
binlog_checksum  NONE  CRC32
 New variable in 5.6. 5.5 doesn't accept the checksums. Use NONE if you want to read binlog with 5.5 or on master if have replication slaves using 5.5.
binlog-row-event-max-size  1k  8k  no binlog_row_event_max_size variable, just the option form.
flush_time  1800  Windows changes from 1800 to 0
 Was already 0 on other platforms
host_cache_size  128  128 + 1 for each of the first 500 max_connections + 1 for every 20 max_connections over 500, capped at 2000
 New variable in 5.6
innodb_autoextend_increment  8  64
 64 is 64 megabytes
 0  8. On 32 bit Windows only, if innodb_buffer_pool_size is greater than 1300M, default is innodb_buffer_pool_size / 128M
innodb_concurrency_tickets  500  5000

innodb_file_per_table  0  1

 5M  48M
 InnoDB will change size to match my.cnf value. Also see innodb_log_compressed_pages and binlog_row_image
innodb_old_blocks_time 0  1000
1 second
innodb_open_files  300  300. If innodb_file_per_table is ON, higher of table_open_cache or 300
 20  300
 0  1
innodb_stats_on_metadata  on  off
join_buffer_size 128k  256k

max_allowed_packet  1M  4M

max_connect_errors  10  100

open_files_limit  0  5000
 See Note 1
query_cache_size  0  1M

query_cache_type  ON/1  OFF/0

sort_buffer_size  2M  256k

 See later post about default my.cnf for STRICT_TRANS_TABLES
sync_master_info  0  10000
 Recommend: master_info_repository=table
sync_relay_log  0  10000

sync_relay_log_info  0  10000
 Recommend: relay_log_info_repository=table. Also see Replication Relay and Status Logs
table_definition_cache  400  400 + table_open_cache / 2, capped at 2000
table_open_cache  400  2000   Also see table_open_cache_instances
 0  8 + max_connections/100, capped at 100

Note 1: In 5.5 there were already rules to ask the OS for the highest of (10 + max_connections + table_cache_size * 2) or (max_connections * 5) or the specified/default open_files_limit. The default is now calculated but the other rules are still used. If the OS refuses to allow as many as requested, max_connections and table_cache_size are reduced and you will see a "Changed limits" warning in the error log, as before.

We are also adding a new default my.cnf file and guided instructions on the key settings to adjust. More on this in a later post. We're providing a page with suggestions for settings to improve backwards compatibility. The old example files like my-huge.cnf are obsolete.

Some of the improvements are present from 5.6.6 and the rest are coming. These are ideas, and until they are in an official GA release, they are subject to change. As part of this work I reviewed every old server setting plus many hundreds of emails of feedback and testing results from inside and outside Oracle's MySQL Support team and the many excellent blog entries and comments from others over the years, including from many MySQL Gurus out there, like Baron, Sheeri, Ronald, Schlomi, Giuseppe and Mark Callaghan.

With these changes we're trying to make it easier to set up the server by adjusting only a few settings that will cause others to be set. This happens only at server startup and only applies to variables where you haven't set a value. You'll see a similar approach used for the Performance Schema. The Gurus don't need this but for many newcomers the defaults will be very useful.

Possibly the most unusual change is the way we vary the setting for innodb_buffer_pool_instances for 32-bit Windows. This is because we've found that DLLs with specified load addresses often fragment the limited four gigabyte 32-bit address space and make it impossible to allocate more than about 1300 megabytes of contiguous address space for the InnoDB buffer pool. The smaller requests for many pools are more likely to succeed.

If the value of innodb_log_file_size or innodb_log_files_in_group changes to be different from the files on disk, you will see a message like this in the error log file at the next restart, instead of the old error message:

[Warning] InnoDB: Resizing redo log from 2*64 to 5*128 pages, LSN=5735153

One of the biggest challenges for the defaults is the millions of installations on a huge range of systems, from point of sale terminals and routers though shared hosting or end user systems and on to major servers with lots of CPU cores, hundreds of gigabytes of RAM and terabytes of fast disk space. Our past defaults were for the smaller systems and these change that to larger shared hosting or shared end user systems, still with a bias towards the smaller end. There is a bias in favour of OLTP workloads, so reporting systems may need more changes. Where there is a conflict between the best settings for benchmarks and normal use, we've favoured production, not benchmarks.

We're very interested in your feedback, comments and suggestions.

2012-10-15 Correction: the initial version of this post wrongly said that innodb_autoextend_increment "Now affects *.ibd files". It still does not and I've updated the table.